AI All The Things
Real fixes from real agent builds. Every article comes from a live session — version-pinned environments, reproduction steps, and fixes that actually ran in production.
Latest
- ai-readiness Don't Make Your Agent Call an LLM to Tell What CMS a Page Is Built On Strong vs weak signals, confidence tiers, fixed iteration order. Heuristic page classification gives your agent a deterministic answer in microseconds.
- general Use AI Like Seasoning, Not Like Flour Deterministic code is the dish. AI is the seasoning. Most of what we reach for an LLM to do, a regex would handle faster and more honestly.
- ai-readiness Your MCP Server Needs Two Tools, Not One Most MCP servers ship one tool that does the work. They should ship two — one that produces a result, one that retrieves it by ID. Here's why.
- claude-code One Claude Code Binary, Ten Different Agents — Shell Aliases as Workflow Modes Shell aliases that set model, effort level, and system prompt turn one Claude Code install into specialized agents for planning, review, coding, and content.
- claude-code Write the Plan with One Agent, Break It with Another — The Review Loop That Catches Real Bugs One Claude session can't review its own work. A second session with a different system prompt finds critical gaps in under 60 seconds.
- claude-code Claude Code's Deny Rules Don't Protect You — Here's What Actually Does Deny rules only block Claude's built-in tools. Bash subprocesses bypass them entirely. Sandbox mode is the real security boundary.
- general Don't Have Claude Do Your Taxes Taxes look like a lookup table. They're actually a two-layer problem — and knowing where to draw the line is the whole skill of building with AI.
- obi-jam Discord Image Attachments → Claude Vision in Five Lines Route Discord image attachments to Claude's vision API. Detect by content type, build image content blocks, send images before text.
- obi-jam Channel-Aware Command Enrichment for Discord Agents Commands in a specific channel get auto-injected with project context. Same command, different channel, different behavior — no user effort required.
- claude-code Official AI Agent Skills Are Here — Install Stripe and Cloudflare Best Practices in One File Companies now ship SKILL.md files that encode their engineering team's best practices. Here's which ones exist and how to install the three that matter most.
- obi-jam Building Per-Skill LLM Cost Tracking Into Your Agent A JSONL-per-month cost logger that tags every API call by skill and user. Know exactly what each capability costs — not just total spend.
- claude-code MCP Tool Schemas Accept Plain Objects — But Only Zod Actually Works The MCP SDK silently accepts plain JSON Schema objects for tool parameters. Inputs arrive as undefined. Only Zod schemas validate at runtime.